Determination of Dissolved Oxygen in Hydrocarbons ANDERSON B. MCKEOWN and R. R. HIBBARD Lewis Flight hopulsion Laboratory, National Advisory Committee for Aeronautics, Cleveland, O h i o
A method is described for the determination of dissolved oxygen in hydrocarbons by a modified Winliler procedure. The procedure is easily run using siniple equipment. Mercaptans do not appreciably affect the results, but peroxides interfere; a correction factor is proposed to compensate for the effects of peroxides. 'The repeatability of the method is good and the accuracy appears comparable to that obtained by more lahorious techniques.
T
HI,: concentration of dissolved oxygen in petroleum fractions i? of interest in connection nith their storage and thermal stability, sweetening processes, and probably other applications. Nost procedures for the determination of dissolved oxygen in organic solvents involve the removal of dissolved gases from the solvtJritj and the subsequent determination of oxygen in these gaws ( 1 , 2, 4 , 8, 11). These procedures generally require special or elaborate apparatus and most of them are time-consuming. The polarograph has been used by Hall ( 7 ) to determine dissolved osygen in fuels; this method is fairly rapid and does not require the .separation of dissolved gases. A simpler method than those mentioned is the modified Winkler procedure proposed by Schulze, Lyon, and Morris (15). In this method the fuel is shaken with an alkaline suspension of manganous hydroxide, the hydrocarbon and aqueous phases are separated, and the aqueous phase is acidified in the presence of iodide. The resultant iodine is titrated with thiosulfate. This method, which has been proposed for the determination of dissolved oxygen in sweetened gasolines, presents considerable difficulty in the separation of hydrocarbon from caustic-manganese oxide suppensions, particularly when applied to fairly viscous fuels, such as kerosine and jet fuels. The method proposed herein is a further modification of the Winkler method, which eliminates the problems of phase separation that are encountered in the Schulze procedure and is not significantly influenced by the mercaptan concentration8 found in most fuels. Fuels are shaken with an alkaline suspension of manganous and ferrous hydroxides and the system is acidified prior to separating the fuel and aqueous phases. The net effect is to form ferric ion in proportion to the amount, of oxygen dissolved in the sample. The amount of ferric ion is determined iodometrically. Both the classical Winkler method and the Schulze procedure use the iodide ion rather than the ferrous ion as a reducing agent. A reducing agent is required when the system is acidified; otherwise the manganic ion would go through a disproportionation reaction to yield the manganous ion and manganese dioxide. The dioxide is not soluble in acids of the strength used in this work and would not be as easily determined as either iodine or the ferric ion. The ferrous-ferric system is used in the proposed method, because it permits acidification in the presence of the fuel. If iodide were used, the resulting iodine would both dissolve in and possibly react with the fuel. For this reason the fuel and aqueous phases must be separated prior to acidification in the Schulze procedure (15). Organic peroxides interfere in the proposed procedure and presumably a similar interference would be found in the Schulze method (16). However, an empirical correction is proposed to minimize errors due to peroxides. The method has been tested on a variety of fuels, solvents, and pure hydrocarbons and is believed to be generally applicable to liquids that are immiscible
Figure 1.
Apparatus used in determination of dissolved oxygen
with water. It should also be applicable to partially or fulljmiscible liquids, provided these do not react with iodine and there is no interference with the starch indicator; however, the method has not been tested on such compounds. The method uses simple apparatus and common reagents. I t requires 1.5 to 2 hours per determination, but several analyses can be run simultaneously. APPARATUS
The fuels were shaken with the reagents in 500-ml. sealing ampoules similar in shape to the bottles in which bromine is sold. These were cylindrical glass bottles about 8 em. in outside diameter and 12 cm. long, with a tubulature 1.0 cm. in outside diameter. Short lengths of Tygon tubings/,,inch in inside diameter were fitted over the tubulature to leave a 5-cm. free length of tubing. The ampoules were closed by using pinchclamps on this free length. The ampoules were shaken with a rocking motion by a Burrell Model C12 shaking machine which had been fitted with cups to hold the ampoules. The ampoules and modified shaking machine are shown in Figure 1. I t is probable that other containers and shaking methods would be equally effective. REAGENTS
Manganous-ferrous solution, containing approximately 0 . 2 s manganous sulfate, 0.2N ferrous ammonium sulfate, and 0 . 2 5 hydrochloric acid. Sodium hydroxide, approximately 2.5A7. Hydrochloric acid, approximately 5.0X. Potassium iodide crystals. Sodium thiosulfate, approximately 0.1A7, standardized. Starch indicator. PROCEDURE
Place 50 ml. of the manganous-ferrous solution in an ampoule and connect via the Tygon tubing to one leg of a three-way stopcock. Connect the other two legs to vacuum and to a source of oxygen-free gas. A vacuum of ca. 5 mm. absolute is adequate and any convenient inert gas may be used; propane was used in this work. Draw a vacuum on the ampoule until the solution boils for a few seconds under the reduced pressure and then admit the inert gas to near atmospheric pressure. Repeat the evacuation and repressurization three times to flush out atmospheric oxygen. Finally evacuate for 1 to 2 minutes and close off the
1490
1491
V O L U M E 2 8 , NO. 9, S E P T E M B E R 1 9 5 6 ampoule with a pinchclamp. Disconnect from the stopcock and, using a small funnel, add 20 ml. of sodium hydroxide without breaking the vacuum. Squeeze out any air bubbles trapped in the tubing before opening the pinchclamp. Leave the last few drops of caustic above the pinchclamp as a precaut'ion against air's leaking int.0 the ampoule. Add 250 to 300 ml. of fuel in the same manner. Place the ampoule in a shaking machine and shake for 30 minutes. Then add 25 ml. of 5N hydrochloric acid and shake manually for a few seconds to neutralize the caustic. Open the pinchclamp and drain the contents of the ampoule into a separatory funnel. Wash the ampoule with 20 ml. of water and add this water to the separatory funnel. Drain the aqueous phase from the separatory funnel into an Erlenmeyer flask, add 5 grams of potassium iodide crystals, allow the solution to stand for 10 minutes, and titrate with O.1N sodium thiosulfate, using starch solution indicator. Record t,he volume of thiosulfate required and then set the flask aside for approximately 30 minutes. If, at this time, color has returned to the starch indicator, titrate again n.ith thiosulfate and add this volume t o the first reading. As the ferrous-manganous-hydrochloric acid solution is slowly oxidized by air, run a blank upon this solution each time it is used, repeating the above procedure in all its steps except the addition of the fuel sample. Measure the volume of sample left in the separatory funnel, and calculate the amount of dissolved oxygen by the following equation: Dissolved oxygen, ml. (STP) per liter = net ml. of Na2S208(sample - blank) X normality X 5603 ml. of sample RESULTS AND DISCUSSION
The repeatability of this procedure is shown in Table I by the results from quadruplicate tests run on seven fuels. One result (for kerosine) was omitted from the average, because the Dean and Dixon rejection coefficient ( 3 ) showed that there was a greater than 90% probability of experimental error. A standard deviation, Q, of 0.67 ml. per liter was found for the method.
Table I. Repeatability of Dissolved Oxygen Determinations with Air-Saturated Fuels Dissolved Oxygen, Ml.(STP)/Liter Av. 3 0 . 8 , 3 3 . 4 , 3 4 . 0 ,3 3 . 0 32.8 40.9,41.0,41.0,40.8 40.9 62.0, 62.4. 61.4,61.2 61.8 46.6, 46.7, 46.4, 46.5 46.6 3 8 . 8 39 0 3 9 . 3 ( 3 7 . 1 ) " 39.0 2 8 . 7 : 28:3: 2 8 . 1 : 2 7 . 9 28.3 41.4, 42.3, 42.7, 42.7 42.3
Fuel Benzene Toluene n-Heptane Cyclohexane Kerosine .Mineral oil, light Clear gasoline
No experiments were run which directly show the sensitivit? of the method. However, data were obtained on two jet fuels which contained relatively small amounts of dissolved oxygen. These were prepared by saturating the fuels with a gas containing 2.0 volume % of oxygen; this gas Tyas made by taking a tank of air a t 1.0 atm. and pressurizing it to 10.5 atm. with cylinder nitrogen. Dissolved oxygen determinations were made on these fuels and also on the same fuels \$-hensaturated with air. The experimental values are shown below. Also shown are the solubilities calculated for the 2% oxygen-saturated solutions using the data on air-saturated fuels and Henry's law.
Sample
Found, Air Saturated
j:
46.8 50.1
Dissolved Oxygen, M1. (STP) /Liter 2% 0 9 Saturated Found Calcd. Difference 4 ,5 4.8
4.3 4.4
0.2 0.4
These data shon that accuracy is preserved down to relatively low concentrations of dissolved oxygen. The differences are well below the standard deviation for air-saturated fuels. Table I1 lists the dissolved oxygen concentration found for various air-saturated hydrocarbons. Because saturation was accomplished by bubbling large excesses of air through the solvent, the dissolved oxygen should be 0.21 that obtained with pure oxygen. Therefore Ostwald coefficients-i e., the volume of gac: at the pressure and temperature of the experiment dissolved in one volume of solvent-for pure oxygen were calculated and are also listed in Table 11. The calculation of these coefficients requires the vapor pressures of the hydrocarbons a t the temperature of the air-saturation experiment and these were taken from the data of Rossini ( 1 4 ) for the pure hydrocarbons and estimated ( 1 2 ) from the Reid vapor pressure for the commercial fuel.
As a n example, benzene, when air-saturated a t 26" C. and 748 mm. total pressure, had a dissolved oxygen content of 32.8 ml. per liter (Table 11)or 0.0328 ml. per ml. of solvent. For pure oxygen 100 the solubility would be 0.0328 X - = 0.156 ml. of oxygen per 21 ml. As the vapor pressure of benzene is 100 mm. a t 26' C. ( 1 4 ) ) the partial pressure of air was 748 100 = 648 mm. Converting the 0.156 ml. of oxygen from standard temperature and pressure t o the conditions of air saturation (26' C. and 648 mm.) gives 0.200 as the Ostwald coefficient for benzene. The vapor pressures of dodecane, hexadecane, and fuels with zero Reid vapor pressure are so low that they need not be considered.
-
.
An attempt was made to estimate the accuracy of the proposed method by comparing data for air-saturated hydrocarbons with literature values for the solubility of oxygen in the same solvents. = 0.67. Comparison was made on the basis of Ostwald coefficients (Table 111). The repeatabilities of both this method and the several methods Droposed in the literature are good, well below 0.01 Table 11. Dissolved Oxygen Content of Air-Saturated Petroleum Fractions and Pure Hydrocarbons unit in Ostwald coefficients. For reasons unReid SaturaSatura- Dissolved known, the differences among the literature GravBoiling Vapor tion tion Oxygen Ostwald Ostwald coefficients for benzene and toluene are PresTemp., Pressure. M I . (STP)/ C,oeffiity, Range, Substance OAPI ' F. surea C. h l m . Hg Liter clents considerably greater than the repeatabilities of Benzene .. ... .. 20 748 32.8 0.200 the several methods. Perfect agreement is shorn-n Toluene .. ... .. 26 751 40.9 0.224 between the literature values for n-hexane and p -X y 1en e ... .. 27 750 38.4 0.206 n-Hexane .. ... .. 23 749 58.0 0,372 iso-octane (6, 6); however, both references are n-Heptane ... .. 24 752 61.8 0,344 Iso-octane .. .. 25 745 64.7 0,368 by the same authors. Only single literature n-Dodecane .. ... .. 25 748 38.3 0.202 n-Hexadecane 750 36.3 0.192 values could be found for n-heptane and cyclo26 Kerosine 4412 340-530 0'0 ' 26 749 3 49 .. 70 0 .. 22 03 58 hexane. Because of spread in some of the literaTarsol 49.2 350-388 26 744 4 0 Light mineral oil 30,4 .., 0 28 748 28.3 0.151 ture Ostwald coefficients, the accuracy ctf the Motor gasoline 60.0 9.0 26 743 30.4 0.305 mcthod cannot be stated, but the comparisons of Furnace oil No. 2 3 5 . 6 3561622 0 24 742 33.2 0.176 100/130 gasoline 71,4 110-328 6.4 22 747 48.8 0.353 T a t l e I11 suggest that the accuracy of the method Clear gasoline 58.7 112-351 5.3 744 42.3 0,306 JP-4 52.9 160-422 2.2 24 753 42.8 0.248 is probably within & l o % of the actual amount. JP-5 42.2 360-502 0 23 745 39.3 0.206 As peroxides readily oxidize the ferrous ion and Measured on another sample after air saturation to eliminate effect of loss of light ends. are often determined by means of this reaction (18), the results of the dissolved oxygen con-
d& a
Value in parentheses omitted from average.
.
.
1
Standard deviation, c, =
ANALYTICAL CHEMISTRY
1432 centrations are certain to be high in the presence of these compounds. If peroxides in fuels reacted completely with the ferrous-manganous reagent, one peroxide number-Le., 1 nieq. per liter-would be equal to 5.6 ml. of oxygen per liter of fuel. It has been shown (18),hnxever, that many peroxides have efficiencies w-ell below 100% in the conversion of ferroiis to ferric ion; therefore a study was made of the effect of peroxides on this dissolved oxygen determination. The fuels used for this study n-ere peroxidized by the action of pure oxygen in the presence of sunlight or by the addition of measured quantities of organic peroxides to the fuels. The results are shown in Table IT’, along with data taken from W:igner, Smith, and Peters (18). These data shon- that osidation of both ferrous and ferroiia-manganous systems by peroxide? was far lese than theoretical: the average oxidation cfficiency of the perosides vias 21TC for the ferrous-manganous systmi and 4 0 5 for the ferrous system. A correction, therefor?, of 0.3 of theoretical is proposed for this method. This correction iq &out ,niidway between t,he above Irentioned average oxidation efficiencies. The correction factor amounts to 5.6 X 0.3 = 1 . 7 nil. of oxygen per liter of fuel per peroxide niimber of the fuel; it should be subtracted from the determined dissolved oxygen rontent. This correction, even though it is empirical and based on limited data, should rarely be in error by more than 2.0 nil. of oxygen per liter of fuel per peroxide number for commercial fiieis. The errors might be slightly larger for pure compounds. The method of Schulze, Lyon, and Morris ( 1 6 )is recommended onl>- for sweetened gasolines, presumably because the mercaptans in unsn-eetened gasolines are extracted by the alkaline manganous-manganic oxide suspension and subsequent1:- oxidized by iodine. This gives low results for dissolved oxygen. A similar effect’ of mercaptans on the results of the proposed procedure a-ould be expected only if the mercaptans remained in t’he aqueous phase after acidification, as the iodide ion is not added until after acidification and phase separation. Tn.0 fuel?, iso-octane
and kerosine, each containing approximately 0.01 weight c> mercaptan sulfur as added n-butyl mercaptan, were studied in order to evaluate the effect of mercaptanq on the results of thiq mi,thod (Table V).
Table
Y. Effect of Mercaptans on Dissolved Oxlgen Determination Dissolved Oxvgen hll /LiterSample
Iso-octane Iso-octane Kerosine Kerosine
I0.01
mercaptan S
+ 0 . 0 1 mercaptan S
Temp., C.
Tlits Work
Benzene Toluene ??-Hexane n-Heptane Iso-octane Cyclohexane
2G
0.200 0 224
2R
23
( 6 1 7 9 )
(10)
o,228
224
o , 174
O ’ i i 2 0.232
... .... .. .
0.179
0.372 0.314 0.368
0.359
24 23
25
0.282
,, ,
0 368
0.359 0.353 0,368 0.222b
, , , ,,,
. .. ... ..,
.,,
(11) , ,
,
.., . ., 0:373
...
64.4, 64.2 63 5 35 3 3i.8
61 7 53 4
39 0 36.4
rz,
LITERATURE CITED
-
Literaturea (6)
65.4, 63.6. 39.0, 35.7,
The data iridi(,ate that mercaptan siilfiu, milcentrations of order of 0.01 w i g h t which are abnormally high for ordinni,y fuek, have but R minor effect on the diqsolred oxygen resii1t.c. There was no signihrant change in the rercaptan concentration of these fuels during the dissolved oxygen determination. It i* believed that no significant inacciiracies due to mercaptans will result when the method is applied to normal petroleum product?: for these products the mercaptan concentration is usually n-ell helow 0.0055. Considerable difficultj- was experienced during the developnient of thiq method, in obtaining permanent end points in the iodornetric determination of t,he ferric ion. S,-ift (16’1has show1 that permanent end points can be obtained for this titration if at least 3 grams of pot,assirim iodide and acid strengths of 0.25 t o 12 meq. of hydrochloric acid in 30 nil. of soltition are used for the iodide reaction. His conditions may not hc applicable to the reactions in this procedure, aF; his solutions did not contain nianganeEe. I t n-as found in the developnient of this procedure that end points are nearly permanent if the acid strength used for the iodide reaction is about 0.5 to 1.0 meq. of hydrochloric acid per milliliter of soliltion. The amoiints of reagents specified in this procedure give final acid concentration? within these limits.
Table 111. Experimental and Literature Ostwald Coefficients Hydtocarbon
.%I.
64.8, 63.2, 38.8, 35.8,
(IS) o,209
... , .. 1:: ...
S.G.. J . Inst. Petroleum 39. 105 11953) ~, . (25 Brooks. F. R . , Dimbat; AI., Treseder, R. S., Lykken, L., r l s a ~ . (?HEX. 24, 520 (1952). (3) Dean, R. B., Dixnn, W. J., Ibid., 23, 636 (1961). (4) Gemant, A . Trans. Faraday Soc. 32,694 (1936). (5) Glendinning, W. G., Bedwell, A I . E.. Royal -%ireraft Estahlish-
(1) Baldwin. R. R.. Daniel.
I
~
ment, Chem. Rept. 464 (October 1949).
(6) I b i d , 477 (February 1951). 6 17’1 Hall. hI. E.. . h A I . . C H E U . 23. 1382 (19611. Hooper, J. H., Proe. Aim Petrheum inst. 28,111, 31 (1948). Horiuti, J., Sei. Papers Irist. Phys. Cheni. Research (Tokyo) 17, 126 (1931). Table IV. Oxidation Efficiencies of Peroxides International Critical Tables, vol. 111,. v. - 263, XcGraw-Hill, Sew York, 1928. This T o r i c ( 1 8 ) Table I Oxidation Oxidation (1 1) Kretschmer, C. B., Towakon-ska, J., TTiehe, R., Dissolved efficiency efficiency Ind. Eng. Chem. 38, 506 (1946). PPTbased on oxygen Perbased on (12) llaxn~ell, J. B., “Data Book on Ilydrocarfound, oxidea dissolved oxide peroxide bons,” p. 44, Van Sostrand, Sew l-ork, Substance ml./ltter So. 02 analysis No. b analysis 1950. ... Cumene 3i.4 1 Gc 9.1 ib 45.9 13:86 73.5 Morgan, J. L. R., Pyne, H. R., J . Phys. Chem. cumene peroxide Cumene 17.5 1.8c Tetralin 34,2045 (1930). 17.8 32 7r92 2 5 . i , ’ 2 7 . 0 46.0 Tetralin peroxide Tetralin Rossini, F. D., others, “Selected Values of Pro0 63.1 Iso-octane 5 . 5 is di-tert-butyl peroxide io 5 Iso-octane perties of Hydrocarbons,” Natl. Bur. StandJP-4 42 5 1 2c .. ards, Circ. C461 (1947). 3.8 8 43.8 JP-4, peroxidized 39 5 JP-5 0 3c .. Schulse, W. A , Lyon, J. P., M o r r i s , L. C., Oil 6 1 25 JP-5. ieroxidized 47.8 ... Gas J . 38, 149 (1940). 8.59 78.7.’79.1 .. ... .. iert-Butyl hydroperoxide Su-ift, E. H.. J . Am. C h e m SOC. 5 1 , 2682 11.59 12.2 Hydrogen peroxide .. ... .. 10.10 1 3 . 6 , 13.9 .. ,.. .. Benzoyl peroxide (1929). 8.40 31.7 ... .. .%scaridole .. Wagner, C. D., Smith, R. H., Peters, E . D., 40 4 v . 21 a
Literature values interpolated t o experimental saturation temperatures. Value questioned by authors ( 6 ) .
+ + +
a Analyeis by method of ( 1 7 ) . b Calculated from d a t a originally presented c -4s received from commercial source.
in terms of grams per liter.
~ A L CHEM. . 19, 976 (1947). (18) Ibid., p. 982. RECEIVED for reriew February 28, 1936. .4ccepted June 2 ,
1956.
